The overall goal of this project is to test the effect of patient-specific computer feedback on physicians' test-ordering practices. The 3 kinds of feedback we propose to study sequentially are: 1) patients' previous test results, 2) estimated probability of each patient's having an abnormal test, and 3) recommendations from the literature. The long-term goal is to reduce health care costs by reducing unnecessary testing. To achieve this goal, a computerized test-ordering system will be developed in a general medicine clinic for the 10 most commonly ordered tests. In the first study, physicians will be randomized into 2 groups. The computer will provide only physicians in the study group with patients' previous test results when tests are ordered at the computer terminal, and the physicians may then retract their order. During this study, physicians in both groups will also be asked by the computer about conditions triggering the order. This information, and other data in the existing computerized medical record system, will be used to derive predictive formulas for abnormal test outcomes. In the second study, again a randomized trial, the computer will give physicians in the study group the predicted probability of an abnormal finding when they order each test. In the last randomized trial, the feedback from the computer for the physicians will be recommendations from the literature about when not to order tests. The effectiveness of each intervention in reducing test ordering will be tested statistically as will factors which may affect the physicians' ordering behavior.